What the numbers actually say
Two countries published AI hiring data this year and both got the same headline: AI hiring is booming. Read one level down and the two datasets describe almost opposite things.
- India's demand is for people who build. Industry projections reported in India's 2026 AI job-market analyses put the country at roughly 3.82 lakh (382,000) AI job postings in 2026, up about 32% from roughly 2.9 lakh (290,000) in 2025.
- India's build-side supply is the tightest number in either dataset. Quess Corp's India AI Workforce Analysis 2026, drawn from 3.5 lakh (350,000) analysed job postings, counts about 9.2 lakh (920,000) AI professionals in the country — but only 2.57 lakh (257,000) of them in core AI roles.
- NASSCOM puts the 2027 requirement close to 1 million AI-skilled professionals, against a currently trained pool of roughly 5-6.5 lakh (500,000-650,000).
- Britain's demand is for people who use. The PwC AI Jobs Barometer 2026, published on 23 June 2026, reports AI user roles up 65.8% in 2025, roughly 66,000 positions, against AI developer roles up about 21%, roughly 2,600 positions.
- That British growth happened inside a shrinking market. Overall UK vacancies across the economy fell 6.6% year on year, while specialist AI postings surged 61%, adding 68,000 roles.
- The pay signals differ too. The UK AI-skills wage premium reached 34.2%, up from 11% in 2024. India's ranges are banded by level rather than by premium.
Put crudely: the same sentence means "we cannot find enough engineers" in Bengaluru and "we are retraining the staff we already have" in London. Those are different problems, and if you are a builder in either market, they imply different moves.
India: 382,000 adverts, and a build-side that cannot keep up
The Indian numbers are the more familiar story, and the more frequently misread one. The projection of 382,000 AI job postings in 2026 is a forecast built on posting trends, up about 32% on the roughly 290,000 recorded in 2025. Alongside it, NASSCOM-BCG data has AI engineer roles growing 67% year on year.
The demand is also geographically and sectorally concentrated in a way that matters if you are deciding where to live. Bengaluru, Hyderabad, Pune and Delhi NCR lead the hiring, with tier-2 cities growing off a smaller base. The leading sectors are BFSI, e-commerce and retail, IT services and global capability centres, and healthcare and pharmaceuticals. That last grouping is not a coincidence: we have written before about how capability centres now account for around a third of India's AI hiring, and they buy production experience rather than potential.
The 920,000 number and why only a quarter of it counts
Quess Corp's India AI Workforce Analysis 2026 is the most useful thing published on the Indian market this year, because it does the one thing the headline numbers never do: it splits the workforce by what people actually do. Across 350,000 analysed job postings, it puts the total at about 920,000 AI professionals, and then divides them.
| Segment | Headcount | Share of the 920,000 | What the job is |
|---|---|---|---|
| Core AI roles | 257,000 | Roughly 28% | Building, training, deploying and running AI systems as the substance of the job |
| AI-embedded roles | 663,000 | Roughly 72% | AI as one component of a wider role — analytics, engineering, product, operations |
| Total AI professionals | 920,000 | 100% | Everyone the analysis counts as working in or around AI |
Now hold that 257,000 against NASSCOM's estimate that India will need close to 1 million AI-skilled professionals by 2027, with a trained pool today of roughly 500,000 to 650,000. The two datasets are not measuring identical populations and should not be subtracted from one another as though they were. But the direction is unambiguous, and it is the sharpest supply gap visible anywhere in either country's data: the segment India needs most is the segment it has least of, and it is the segment that takes longest to produce.
That is why 67% growth in AI engineer roles has not made the Indian market feel easy from the inside. Demand is growing fastest exactly where supply is thinnest, and the people who can genuinely fill core roles are, by definition, already employed doing them.
An "AI" job title is now weak evidence in India. If 663,000 of the 920,000 people counted as AI professionals sit in AI-embedded roles, then having AI, ML or GenAI in your designation puts you in the majority group rather than the scarce one. Hiring managers know the ratio, and they have adjusted: the screen has moved from what your title says to what you can show you built and ran. Assume your title is discounted before you have opened your mouth.
Britain: a 61% surge that is mostly people learning to use it
The UK figures come from a single, well-audited source, which makes them easier to reason about. The PwC AI Jobs Barometer 2026, published on 23 June 2026, reports that specialist AI job postings surged 61% in 2025, adding 68,000 roles, and that specialist AI jobs are now 2.2% of the overall UK job market, up from 1.3% the previous year. Taken alone, that reads as a boom.
Then PwC splits the 68,000 the way Quess splits the Indian workforce, and the picture changes. AI user roles — professionals who use AI within their existing work — increased 65.8%, roughly 66,000 positions. AI developer roles expanded about 21%, roughly 2,600 positions. The ratio of new user roles to new developer roles is approximately 25 to 1.
The sector data reinforces it. AI postings by sector run to 10% in technology, media and telecommunications, 8% in health, and nearly 5% in government — the last of those up from under 1% in 2022, which is a genuinely rapid adoption curve for the public sector. But 97% of government AI roles are AI-user positions. Whitehall is not staffing up model teams; it is asking existing civil servants to work with AI tools.
The posting series is a recovery, not a breakout
There is a second correction hiding in PwC's own numbers, and it is the reason to be careful with the word "surge". Look at the UK AI-related posting count across four years.
| Year | UK AI-related job postings | Read |
|---|---|---|
| 2022 | 182,000 | The high-water mark of the series |
| 2023 | 119,000 | A sharp contraction, down roughly a third |
| 2024 | 112,000 | The floor — still falling, more slowly |
| 2025 | Roughly 180,000 | Back to approximately where the series began in 2022 |
The 2025 rebound is real and it is large. It is also, on PwC's own series, a return to the 2022 level rather than an advance beyond it. Three years of AI product releases, capital and public attention have brought UK AI-related postings back to roughly where they were before any of it happened. That context does not make the 61% specialist increase less true; it makes it a recovery within a labour market where overall vacancies fell 6.6% year on year.
We made a related argument in our piece on how the AI share of British hiring hit a record inside a shrinking market. This is the composition version of the same caution: a share can rise because the thing you are measuring grew, or because everything around it shrank, and PwC's four-year series shows a good deal of both.
Everything on this page counts postings, not hires. A single opening can be advertised repeatedly and across several boards; some adverts are never filled; some exist to test the market. Postings are an excellent read on the direction and the composition of employer demand and a poor read on net new headcount. So 382,000 Indian adverts is not 382,000 Indian jobs, and 2,600 net new UK developer roles may correspond to fewer people actually starting one. Use these numbers to understand shape, never to size your own odds.
The same headline, two different shapes
Set the two markets side by side on the six measures that matter and the divergence is easier to hold in your head than any single statistic.
| Measure | India | United Kingdom | What it means for you |
|---|---|---|---|
| Total AI postings | Roughly 382,000 projected for 2026, up from roughly 290,000 in 2025 | Roughly 180,000 in 2025, against 182,000 in 2022 | Advert volume is the same order of magnitude. The difference is not how many, it is what they ask for. |
| Year-on-year change | About +32% projected for 2026; NASSCOM-BCG has AI engineer roles up 67% year on year | Specialist AI postings +61% in 2025, adding 68,000 roles, while overall UK vacancies fell 6.6% | Both are growing. Only one is growing inside an economy that is shedding vacancies elsewhere. |
| Core / developer roles | 257,000 core AI roles of 920,000 AI professionals (Quess Corp) | AI developer roles up about 21%, roughly 2,600 positions added in 2025 | India's build-side is large and short-handed. Britain's added roughly 2,600 net new developer openings in a year. |
| User / embedded roles | 663,000 AI-embedded roles of 920,000 (Quess Corp) | AI user roles up 65.8%, roughly 66,000 positions; government AI roles are 97% user roles | In both markets the majority of "AI headcount" is people using AI inside another job. In the UK's 2025 intake the ratio is about 25 to 1. |
| Trained supply | Close to 1 million AI-skilled professionals needed by 2027 against a trained pool of roughly 500,000-650,000 (NASSCOM) | No trained-supply figure in this dataset — PwC measures postings. The nearest proxy is specialist AI jobs reaching 2.2% of the overall job market, up from 1.3% | India has a published, quantified shortage. Britain has a small specialist slice of a large market and no equivalent supply count. |
| Wage signal | Aggregated market ranges: freshers roughly Rs 6-9 LPA; 3-6 years roughly Rs 12-30 LPA; 7-10+ years roughly Rs 30-60 LPA; GenAI specialists roughly Rs 25-50 LPA | AI-skills wage premium of 34.2%, up from 11% in 2024 | India prices the level you are at. Britain prices the skill you can prove. Both reward evidence over designation. |
One qualifier before anyone builds a strategy on this table. The Indian and British figures come from different organisations using different definitions of what counts as an AI role, over different reference periods. They are not methodologically interchangeable, and the right way to read the table is as two internally consistent stories placed next to each other, not as a single normalised series. Separately, Indeed Hiring Lab reports UK job postings referencing AI at 127% above pre-pandemic levels by February 2026 — a third measurement, on a third basis, pointing the same direction as PwC on adoption.
In a market where hirers search instead of advertise, a profile is not optional
Britain added roughly 2,600 net new AI developer roles last year, and India's shortage sits in exactly the segment where titles prove nothing. Both facts point the same way: the people hiring are searching for evidence, not waiting for applications. AI Tech Connect lists AI engineers, founders and researchers across India and the UK, and the people hiring browse it to find them. It takes two minutes, there is no CV and no password, and it is free.
Claim your Verified Builder profile →What this changes for a builder in each market
The divergence is only useful if it changes what you do on Monday. It should, and differently depending on which side of the comparison you are standing on.
If you build in India
The scarce credential is demonstrable core work, not an AI-adjacent job title. That follows directly from the Quess split: with 663,000 of 920,000 AI professionals in AI-embedded roles, a designation containing "AI" places you in the crowded 72%, not the scarce 28%. The gap NASSCOM is describing — close to 1 million needed by 2027 against 500,000 to 650,000 trained — is a gap in core capability, and no title closes it.
Practically, that means your public record needs to establish which side of the split you sit on before anyone asks. Not "AI/ML Engineer at a large BFSI employer" but the specific systems: what you trained or fine-tuned, what you deployed, what it serves, what broke and what you did about it. Two entries with that level of specificity do more than a decade of job titles, because they answer the exact question the ratio makes hirers ask. Our guide to portfolio projects that actually get people hired covers what that record should contain.
Second, use the salary bands as bands, not as targets. Freshers cluster around Rs 6-9 LPA, three-to-six years around Rs 12-30 LPA, seven-to-ten-plus around Rs 30-60 LPA, with GenAI specialists roughly Rs 25-50 LPA — aggregated market ranges rather than a single authoritative published source, and wide enough that where you sit within one is decided by evidence rather than by years served.
If you build in the UK
The developer market is small and highly competitive, and the honest reading of "roughly 2,600 net new developer positions in a year" is that being findable matters more than applying more. Volume-applying into a pool that size is close to a rounding error on your own time. That is not a counsel of despair — 68,000 new specialist AI roles is a lot of hiring — but it does mean the entrance you are queueing at is narrow, while the adjacent one marked "AI user" is wide open and growing at 65.8%.
There are two rational responses, and they are not in conflict. The first is to compete for the small developer pool on evidence rather than volume, which means being visible to a hiring manager running a search rather than being one of 400 CVs. The second is to notice that 66,000 new user roles are being filled by organisations that mostly do not yet have anyone who can build, and that being the person in a user-heavy team who can actually ship a system is an unusually strong position. Government is the clearest instance: nearly 5% of AI postings, up from under 1% in 2022, at 97% user roles.
On pay, the 34.2% premium is the price of provable skill, not of a job title. It rose from 11% in 2024, which is a fast repricing, and it is a market-level average rather than a number to quote across a table. Treat it as context for a benchmark — our walkthrough of negotiating in a two-tier AI market sets out how to do that without anchoring on somebody else's average. It also helps to know what a hiring manager's rubric is actually scoring before you decide what to put in front of one.
Write your record for the side of the split you want to be on, in either country. For a core or developer role that means naming the system, the scale and the outcome: "built and run the retrieval layer behind a support assistant over 400,000 documents, own the evaluation set and the rollback path" rather than "worked on GenAI initiatives". The first sentence answers the question both markets are actually asking — can this person build one? The second one is the sentence 663,000 people in AI-embedded roles have already written.
Being findable is the part you control
Strip both datasets back and one mechanism is doing most of the work. India's shortage is in a segment where job titles have stopped distinguishing anyone, so hirers screen on evidence. Britain's developer openings are few enough that hirers do not need to run an open funnel, so they search. In neither market does the marginal application do much. In both, a public, specific, findable record of what you have built does.
That is the honest case, and it comes with limits worth stating. A profile is not a job offer. It cannot manufacture experience you do not have, and no directory can promise you a hiring manager's attention. What it removes is one particular failure — being genuinely capable in a market that reports it cannot find capable people, and being invisible to the search that would have found you. Where you have no network to lean on, our guide to getting referred into an AI team without one covers the other half of the same problem.
Both national datasets, in the end, say the same thing about individuals even while they disagree about markets. India is short of people who can build. Britain is short of openings for people who can build, and long on organisations that will shortly need one. The person who benefits from either condition is the one whose work is legible to a stranger running a search. NASSCOM's own community publishes a good deal of the underlying analysis at community.nasscom.in, and it is worth reading the composition tables rather than the headlines.